Step asymmetry post-stroke (i.e., limp) substantially affects the quality of life of stroke survivors because it impairs patients' mobility, thereby limiing their daily activities and increasing their dependency on others. Consequently, a primary interest for patients, clinicians, and researchers is to correct the step asymmetry in stroke survivors. Promising studies show that patients can re-learn to walk symmetrically if their step asymmetry is exaggerated with a split-belt treadmill that moves the legs at different speeds. While these results are encouraging, gait improvements are highly contextual and do not persist when walking over ground. To address this critical issue for gait rehabilitation, the PI is proposing a combination of computational and experimental approaches to identify key factors regulating the generalization of locomotor learning after stroke. The PI's central hypothesis is that inherent features from one's movement (e.g., kinematic errors and walking speed) regulate the generalization of locomotor learning. This hypothesis was formulated on the basis of the PI's preliminary data showing more generalization of treadmill learning to over ground walking when kinematic errors or walking speed during split-belt adaptation are similar to those naturally experienced. In the proposed computational approach, model inputs are errors that subjects experience during split-belt walking (for example, unexpected leg motions disturbing one's balance), model outputs are actions to correct these errors (for example, a larger step to prevent falling). The mathematical relationship between inputs and outputs is used to predict the effect of error size (Aim 1) and walking speed (Aim 2) on the generalization of learning in an individual basis. Once factors mediating the generalization of learning are identified, they can be harnessed to develop interventions that improve the gait of stroke survivors during real-life situations. PI qualifications: the PI is a prolific and creative bioengineer. Her first class trainng in physics, biomechanics, and neuroscience, in addition to her strong interest in rehabilitation make her the adequate individual for doing the proposed work. Her studies in human motor control are well recognized (>700 citations; h-index 11) in a relatively short, but highly productive academic career. Through this award she will receive mentorship from two extraordinary investigators with complementary expertise: Michael Boninger, MD, PhD. (clinical rehabilitation) and Reza Shadmehr, PhD (computational motor control). They will serve as primary co-mentors. In addition the PI will receive mentorship from an expert panel of collaborators including Dr. Subashan Perera (biostatistics), Dr. Steven Graham (neurology), Dr. Julie Fiez (neuropsychology) and Dr. Skidmore (post-stroke rehabilitation). Thus, this award will provide the mentorship and career development allowing the PI to become an independent researcher able to compete for R01-level funding to study gait deficits post-stroke through computational modeling.